Configurable approvals for an AI logistics agent: what runs alone and what waits for a person
Key Takeaways
Updated November 4, 2026
Configurable approvals decide which freight actions an AI logistics agent can finish on its own and which ones wait for a named person. The company sets the rules: rate limits, the approved carrier list, rebooking permissions, and payment thresholds. Wilson, by Cartage, is an AI logistics coordinator that starts with assisted execution, works only inside those rules, and sends anything outside them to the right person with the facts attached.
- An approval rule has four parts: the action, the limit, the approver, and what happens if nobody answers.
- Wilson, by Cartage, books only carriers the company has approved. Cartage can present additional capacity, but the company approves each carrier before Wilson uses it.
- Wilson negotiates rates against the company’s rate targets. Rates outside the approved limits go to the logistics team.
- Rebooking after a missed pickup runs within the rules or with approval, depending on how the company configures it.
- Wilson audits each carrier invoice against the POD and the rate confirmation. Payments above approval limits or outside the rules wait for a person, and there are no partial payments.
- Every action and approval is saved in a searchable, exportable shipment history with change history.
- Teams usually start with more approvals and remove them lane by lane as the work proves reliable.
Which freight decisions need an approval rule?
Freight decisions need an approval rule when they commit money, change who moves the load, or change what the customer was promised. An AI logistics coordinator can do the asking, checking, and recording for every one of them. The rule decides whether it can also finish the step. The table shows a common starting point. It is a hypothetical example, and each company sets its own values.
| Decision | Typical starting rule | Who approves when the rule is exceeded |
|---|---|---|
| Accept a carrier quote | Auto-accept at or under the lane rate target | Logistics manager |
| Counter-offer a carrier | Negotiate up to a set percentage over target | Logistics manager |
| Use a carrier | Only carriers on the approved list | Transportation lead approves new carriers |
| Rebook a missed pickup | Rebook from approved carriers within the rate limit | Shipment owner |
| Change pickup or delivery date | Accept if the customer’s date still holds | Customer service owner |
| Accept an accessorial charge | Accept documented charges under a dollar limit | Freight payables owner |
| Pay a carrier invoice | Pay when the invoice matches the rate confirmation and POD | Finance approver |
| Send a customer update | Send routine status updates automatically | Account owner for bad news or date changes |
The pattern in the table is simple. Routine, rule-checked actions can run without a person. Actions that spend more, add risk, or change a promise go to a named owner.
Three types of decisions almost always deserve a rule. The first is money: quote acceptance, counter-offers, accessorials, and payment. The second is the carrier: who is allowed to move the freight. The third is the promise: dates and details the customer already relies on.
Decisions that rarely need an approval are the ones that only gather or share facts. Asking a carrier for a quote, requesting a pickup window, chasing a missing POD, and posting a status update to the team can usually run within the rules from day one. Those tasks take most of a coordinator’s day, and they do not commit the company to anything new.
Approval is not the same as visibility
Some teams confuse an approval with a notification. A notification tells a person that something happened. An approval stops the action until a person decides. Both are useful, but a buyer should ask which one a tool means. An AI logistics coordinator that only notifies after it books an expensive carrier has no real approval control.
Who should own each approval
Each approval needs one named owner and one backup. Ownership by role works better than ownership by person, because people change shifts, take vacations, and move jobs. A hypothetical manufacturer might give rate approvals to the logistics manager, new carriers to the transportation lead, date changes to customer service, and payments above the limit to accounts payable. The AI logistics coordinator then knows exactly where each decision goes, and nobody has to forward requests by hand.
Avoid approvals that belong to a group. When a decision goes to a shared channel with ten people, each person assumes someone else will answer. A single owner with a clear backup is faster and easier to audit.
Approvals that slow teams down
Too many approvals can be as costly as too few. When every quote under target still needs a manager’s reply, loads wait in a queue and carriers move their trucks to other shippers. The manager also starts approving without reading, which removes the control the approval was meant to add. A useful sign is the approval rate: if a type of request is approved almost every time with no changes, the rule is probably too tight. Moving that decision inside the rules frees the approver for the cases that need judgment, such as a new lane, a large rate increase, or a customer date at risk.
How configurable approvals work, step by step
Configurable approvals work as a short loop: the AI logistics coordinator prepares the decision, checks it against the rule, and either finishes it or routes it to the approver. The loop below shows the steps for a rate decision. Other decisions follow the same shape.
Step 1: The agent gathers the facts
Before any decision, the AI logistics coordinator collects what the approver would need. For a rate, that means the lane, the quotes received, the rate target, the carrier’s recent responses, and the required pickup date. Without these facts, an approval request becomes a new research task for the manager.
Step 2: The agent checks the rule
The rule answers one question: is this action inside the limit? A quote at or under target passes. A quote above target but within the negotiation limit leads to a counter-offer. A quote above the limit, or from a carrier not on the approved list, cannot be accepted without a person.
Step 3: The agent finishes or routes
If the action passes, the AI logistics coordinator completes it and records why it passed. If it fails, the agent sends the decision to the named approver with a short summary: what is being decided, the options, the recommended option, and the deadline that matters.
Step 4: The approver decides
The approver accepts, rejects, or changes the proposal. The decision should take one message, not a meeting. A good approval request makes the answer obvious or at least quick.
Step 5: The agent records and continues
After the decision, the AI logistics coordinator carries it out and saves who approved what, when, and on which facts. That record is what lets a manager trust the next automatic action, and it is what an auditor or a finance team will look for later.
This loop keeps the person in the decision and out of the busywork. The manager does not chase quotes or copy rates into a spreadsheet. The manager reads one summary and answers it. For more on where rate limits sit inside a negotiation, see how much AI can negotiate before a person approves the rate.
A hypothetical example
A hypothetical distributor sets a rate target of $1,900 for a Dallas to Atlanta truckload and allows counter-offers up to 5% over target. Three approved carriers quote $1,850, $2,050, and $2,200. The AI logistics coordinator accepts the $1,850 quote because it is under target, books the load, and records the reason. On the next load, the lowest quote is $2,080. The agent counters at $1,995, inside the limit. The carrier holds at $2,080, so the agent sends the decision to the logistics manager with the three quotes, the pickup date, and the last accepted rate on the lane. The manager approves in one reply.
What a good approval request contains
A good approval request answers the approver’s questions before they are asked. It should fit on one screen and include:
- The shipment and lane, with the pickup date.
- The decision being asked, in one sentence.
- The options, such as the three quotes received, with the carrier names from the approved list.
- The rule that was exceeded and by how much.
- The recommended option and the reason.
- The deadline for the decision and what happens if it passes.
If an approver regularly asks for more information before deciding, the request format is missing something. Fixing the format is often the fastest way to speed up approvals.
Setting the limits: rates, carriers, rebooking, and payment
Approval limits should match the risk of each decision, not one global setting. An AI logistics coordinator can follow a different rule for rates, carriers, rebooking, accessorials, and payment. Setting each one separately keeps routine work moving and puts attention where it matters.
Rate limits
Rate limits usually start from the company’s rate target for each lane or mode. A common setup is: accept at or under target, negotiate within a set range above target, and send anything higher to a person. Wilson, by Cartage, negotiates against the customer’s rate targets under these rules. Rates outside the limits go to the logistics team. Limits can differ by mode, lane, customer, or urgency, because a hot load and a routine replenishment order rarely deserve the same ceiling.
Carrier limits
The safest carrier rule is a closed list. Wilson books only carriers the company has approved and tenders to them first. Cartage can present additional carrier capacity when the approved list cannot cover a load, but the company must approve each new carrier before Wilson uses it. That keeps insurance, safety, and contract decisions with the company.
Rebooking limits
A missed pickup or a carrier cancellation is where speed matters most. Wilson detects the problem, contacts the carrier, notifies the internal team, notifies the customer if configured, and initiates rebooking from the approved carrier list. The company decides whether that rebooking runs automatically within the rules or waits for approval. Many teams allow automatic rebooking when the new rate is within the lane limit and require approval when it is not.
Accessorial and detention limits
Accessorials are small charges that add up. A practical rule accepts documented charges under a set amount and routes the rest. For detention, Wilson audits each detention line against the recorded arrival and departure times and the agreed terms, and sends unsupported charges back to the carrier with evidence. That matters because detention is common: ATRI research published in September 2024 found detention at 39.3% of truck stops in 2023.
Payment limits
Payment rules protect cash. Wilson audits and reconciles each carrier invoice against what was booked, the POD, and the rate confirmation before it pays the carrier from the Wilson Wallet. Payments above the approval limits or outside the rules wait for a person. There are no partial payments: if any invoice line does not match, the whole invoice waits for a person. Companies can also require a signed POD before payment is released.
Document and record changes
Some changes look clerical but carry risk. When a bill of lading changes the address, freight class, carrier, or price, Wilson sends the change to the load owner instead of accepting it silently. Each version is saved with its source and time.
How to write a rule people can follow
Good approval rules are short and specific. “Get approval for expensive loads” is not a rule, because nobody knows what expensive means. “Get approval when the rate is more than 5% above the lane target” is a rule. Each rule should name the action, the limit, the owner, the backup, and the deadline. Rules should also be reviewed after the first month. Limits that trigger approvals that are always granted can usually be raised. Limits that never trigger may be too loose. The aim is not fewer approvals or more approvals. The aim is approvals on the decisions that actually need a person, and no wasted time on the rest.
What happens while an approval waits
While an approval waits, the AI logistics coordinator should keep the load safe without overstepping the rule. A waiting approval is a common failure point: the quote expires, the carrier gives the truck to another shipper, or the pickup date passes. The rule should say what happens during the wait, not only who approves.
Set a decision deadline
Every approval request should carry the time by which a decision is needed. For a quote, that is often the carrier’s validity window. For a rebooking, it is the time the new truck must be confirmed to make the pickup. The deadline turns a vague request into a clear ask.
Escalate when nobody answers
If the named approver does not answer before the deadline, the request should move to a backup approver. The company decides the order. A hypothetical setup sends a rate decision to the logistics manager, then to the director of logistics after 30 minutes on a same-day load. Without a backup, a single person on vacation can stop freight.
Keep the carrier informed
While the company decides, the AI logistics coordinator can ask the carrier to hold the quote or the equipment, where the carrier allows it. It does not accept anything the rule has not cleared. Keeping the carrier informed reduces the chance that the approved option is gone by the time the approval arrives.
Do not stop the rest of the work
One pending approval should not freeze the shipment. The AI logistics coordinator can keep collecting documents, confirming dock hours, and updating the team while a single decision waits. Only the action that needs approval should pause.
Tell the right people
Wilson updates the internal team in Slack, Microsoft Teams, or email, and the customer by email or SMS where configured. A useful practice is to tell the customer only when the decision affects their date, not every time an internal approval opens.
Common waiting mistakes
- Sending approval requests to a shared inbox where nobody owns them.
- Leaving out the deadline, so the approver treats an urgent load like a routine one.
- Asking for approval without the options, which forces the approver to research.
- Letting the quote expire without telling anyone, then starting the search again.
- Approving in a side conversation that never reaches the shipment record.
A hypothetical missed-pickup approval
A hypothetical manufacturer has a truckload scheduled for 8 a.m. The carrier reports at 7:15 a.m. that the truck broke down. The AI logistics coordinator confirms the cancellation with the carrier, notifies the shipment owner in Microsoft Teams, and asks two other approved carriers for same-day capacity. One carrier can pick up at 1 p.m. at a rate 12% above the lane target. The rule allows automatic rebooking up to 8% above target, so the agent sends the decision to the shipment owner with the new pickup time, the rate, and the customer’s delivery date. The owner does not answer in 20 minutes, so the request moves to the logistics manager, who approves. The agent books the carrier, updates the customer by email because the configured rule says date changes are shared, and saves the full sequence in the shipment history.
How to test approval controls before you trust them
Approval controls should be tested on real loads before the team relies on them. An AI logistics coordinator can describe its rules in a demo, but the useful test is to break those rules on purpose and watch what happens. The checklist below takes about an hour with one realistic lane.
| Test | What to do | What a good result looks like |
|---|---|---|
| Over-limit quote | Have the lowest quote come in above the rate limit | The agent counters within the range, then routes to the named approver with options |
| Unapproved carrier | Offer a lower quote from a carrier not on the list | The agent does not book it and asks whether the company wants to approve the carrier |
| Silent approver | Do not answer the approval request | The request escalates to the backup approver before the deadline |
| Missed pickup | Report a missed pickup on a booked load | The agent contacts the carrier, alerts the team, and rebooks within the rules or asks for approval |
| Mismatched invoice | Send an invoice with an extra unsupported charge | The whole invoice waits for a person, and the mismatch is explained |
| Changed BOL | Send a BOL with a different freight class | The change goes to the load owner before it is accepted |
| Record check | Open the shipment history afterward | Every action, approver, time, and reason is visible and exportable |
Score each test as handled within the rules, routed correctly, or failed. A tool that books an unapproved carrier or pays part of a disputed invoice fails the test, no matter how fast it is.
Questions to ask the vendor
- Can limits differ by lane, mode, customer, or urgency?
- Who can change a rule, and is the change itself recorded?
- What happens when the approver does not answer?
- Can the company start with approvals on everything and remove them gradually?
- Can we export the approval history for finance or audit?
How to move from assisted to automatic
Most teams start with assisted execution: the AI logistics coordinator prepares each decision and a person approves it. After a few weeks of matching decisions on one lane, the team can let routine actions on that lane run automatically within the rules. The same process repeats lane by lane. This keeps the decision about trust with the people who run the freight, based on their own results. Cartage does not run pilots; most customers start Wilson on a few lanes and expand from there.
For the record side of this test, see whether freight software can show who approved every change to a load.
Red flags in a demo
- The vendor can show approvals only in slides, not on a live load.
- Limits are global, with no way to set them by lane or mode.
- Approvals happen in email threads that never reach the shipment record.
- The tool can add carriers on its own without the company approving them.
- Invoices can be partly paid while a disputed line is still open.
How Wilson, by Cartage, handles approvals
Wilson, by Cartage, is an AI logistics coordinator that executes freight work inside approval rules the company configures. It runs the full cycle: quote, book, schedule, track, deliver, audit, and pay, across LTL, FTL, ocean, drayage, parcel, and cross-border freight. Wilson replaces the TMS for freight execution and connects to the company’s ERP. At each step, the rules decide whether Wilson finishes the action or routes it to a person.
Wilson starts with assisted execution. Approval limits are set by the customer. Routine actions can run automatically within those rules as reliability is proven. Everything Wilson does is saved in searchable, exportable records with change history.
| Area | What Wilson does within the rules | What waits for a person |
|---|---|---|
| Quotes and rates | Requests quotes from approved carriers, negotiates against rate targets | Rates outside the approved limits |
| Carriers | Tenders to approved carriers first | Any carrier the company has not approved |
| Exceptions | Detects missed pickups, cancellations, and delays, contacts the carrier, and initiates rebooking from approved carriers | Rebooking that needs approval under the configured rules |
| Documents | Prepares and sends rate confirmations, bills of lading, packing slips, and pallet labels; sends the final signed BOL to the customer | BOL changes to address, class, carrier, or price |
| Invoices and payment | Audits each invoice against the POD and rate confirmation, pays carriers from the Wilson Wallet | Payments above limits or outside the rules; any invoice with a mismatched line |
Wilson communicates with carriers by email, API, EDI where supported, and phone where configured. It contacts drivers directly by phone or text when the carrier shares driver contact information. It keeps the internal team updated in Slack, Microsoft Teams, or email, and customers by email or SMS where configured.
Wilson works with SAP, NetSuite, Microsoft Dynamics 365, Oracle, Infor, Epicor, Sage, Acumatica, and other ERP environments, with the exact connection depending on the customer’s data flow and scope. Timing depends on scope: most teams start on a few lanes and expand from there.
Wilson has no subscription or platform fee. Pricing is a percentage of the freight it coordinates, with pricing scaling down as volume grows. Enterprise pricing is custom.
For the broader definition of the category, see what an AI logistics agent is.
What the approval record shows
Each Wilson shipment record shows what was decided, by whom, when, and on which facts. For a rate approval, that includes the quotes, the target, the limit that was exceeded, the approver, and the final booked rate. For a payment, it includes the invoice, the matched POD and rate confirmation, and any line that waited for review. The record can be searched and exported, so finance and operations can review approvals by lane, approver, or month. That history is also how a team decides which approvals to remove next: if a type of decision was approved the same way twenty times, it may be ready to run within the rules.
FAQs
Which AI logistics agents let us configure approval rules?
Wilson, by Cartage, is an AI logistics coordinator that works inside approval rules the company sets for rates, carriers, rebooking, and payment. It books only approved carriers, negotiates against rate targets, and sends anything outside the limits to a named person. Every action and approval is saved with change history.
Can an AI logistics agent pay carriers without approval?
An AI logistics coordinator like Wilson pays carriers from the Wilson Wallet only after it audits the invoice against the POD and rate confirmation and the payment fits the company’s rules. Payments above approval limits or outside the rules wait for a person. If any line does not match, the whole invoice waits.
Can we approve every action at the start?
Yes. Wilson, the AI logistics coordinator, starts with assisted execution, so the team can approve every quote, booking, and payment at first. Routine actions can then run automatically within the rules, lane by lane, as the team sees reliable results.
What happens if the approver does not answer?
The request should escalate to a backup approver before the decision deadline. Your AI logistics coordinator should keep the rest of the shipment work moving, ask the carrier to hold the quote where possible, and record who decided and when.
Sources
- Cartage, Wilson product FAQ and product scope, November 2026: https://cartage.ai
- American Transportation Research Institute (ATRI), driver detention research, September 2024: https://truckingresearch.org/2024/09/new-research-documents-substantial-financial-and-safety-impacts-from-truck-driver-detention/
Want to see your own approval rules running on real loads? Talk to the Cartage team and walk through how Wilson, the AI logistics coordinator, would handle your rate, carrier, and payment limits.